Analysis of the Sensitivity of Algorithms for Assessing the Harmful Information Indicators in the Interests of Cyber-Physical Security

Abstract

The secure functioning of cyber-physical systems depends on the presence and amount of harmful (unwanted and malicious) information in its digital network content. The functioning of cyber-physical systems is carried out in non-stationary conditions and in conditions of continuous exposures. This leads to the uncertainty of indicators (parameters, features) of harmful information that must be assessed in the analytical processing of digital network content. The paper proposes an approach to analyse the sensitivity of algorithms for estimating the status of indicators of harmful information observed in noise. This approach allows one to consider possible errors in the estimation accuracy. It gives the possibility to identify the allowable range of changes in the parameters of the digital network content of cyber-physical systems, within which the requirements for the assessment reliability are met. This, in turn, makes a significant contribution to the effectiveness of harmful information detection and counteraction against it. Accounting for a priori uncertainty of the indicators under various influences is advisable to carry out on the basis of expressions for the sensitivity coefficients (functions) described in the paper.